Telecom Customer Churn Prediction Analysis

"Did you know that attracting a new customer costs five times as much as keeping an existing one?"

A comprehensive data mining project analyzing and predicting customer churn in the telecommunications industry, where annual churn rates reach 15-25%.

👥 Project Team

• Azami Hassani Adnane
• Chegdati Chouaib
• Bellmir Yahya
• Amcassou Hanane
• Benakka Zaid
• Lamkharbech Issa

Supervised By: Pr. Hosni

📌 Introduction

Customer churn refers to when customers discontinue their services with a company. In the highly competitive telecom industry, customers can easily switch between providers, leading to significant customer turnover. This project focuses on predicting potential churners to help companies implement targeted retention strategies.

Why It Matters

• Retaining existing customers is 5x cheaper than acquiring new ones
• High annual churn rates (15-25%) significantly impact profitability
• Predictive analytics can help focus retention efforts on high-risk customers

🎯 Objectives

1. Analyze churn rate distribution
2. Identify gender-based churn patterns
3. Understand service preferences of churning customers
4. Determine profitable service types
5. Build predictive models for churn detection

📊 Data Analysis

Dataset Overview

The dataset includes:

• Customer demographics (gender, age, partners, dependents)
• Service subscriptions (phone, internet, security, streaming)
• Account information (tenure, contract type, payment method)
• Billing details (monthly charges, total charges)

Key Findings

1. Customer Demographics

• Gender distribution: 49.5% female, 50.5% male
• Churn rate: 26.6% of customers switched providers
• New customers show higher churn probability

2. Contract Analysis

• Month-to-month contracts: 75% churn rate
• One-year contracts: 13% churn rate
• Two-year contracts: Only 3% churn rate

3. Payment Methods

Distribution and churn impact:

Payment Method	Share (%)	Risk Level
Electronic Check	33.6%	Highest churn
Mailed Check	22.8%	Moderate
Bank Transfer	21.9%	Low
Credit Card	21.6%	Lowest churn

4. Service Analysis

• Fiber optic users show higher churn rates despite popularity
• DSL users demonstrate better retention
• Higher monthly charges correlate with increased churn probability

🤖 Machine Learning Models

Model Performance Comparison

Model Accuracy:

• Voting Classifier 81.7%
• Random Forest 81.4%
• Logistic Regression 80.9%
• Gradient Boosting 80.8%
• SVM 80.8%
• KNN 77.5%
• Decision Tree 72.5%

Best Model: Voting Classifier

• Ensemble approach combining:
  • Gradient Boosting
  • Logistic Regression
  • AdaBoost
• Achieved highest accuracy: 81.7%
• Balanced precision and recall metrics

📈 Business Recommendations

1. Contract Strategy

   • Encourage longer-term contracts
   • Offer incentives for contract upgrades
   • Design special packages for month-to-month customers

2. Payment Method Optimization

   • Promote automatic payment methods
   • Provide discounts for bank transfer/credit card payments
   • Review electronic check payment process

3. Service Quality

   • Investigate Fiber optic service issues
   • Enhance technical support
   • Implement proactive maintenance

4. Customer Retention

   • Focus on early tenure customers
   • Develop loyalty programs
   • Regular service satisfaction surveys

🛠 Technologies Used

• Python
• Pandas & NumPy
• Scikit-learn
• Matplotlib & Seaborn
• Plotly
• Google Colab

💡 Conclusion

Customer churn significantly impacts business profitability. Our analysis shows that successful churn prevention requires:

• Understanding customer behavior patterns
• Identifying high-risk segments
• Implementing targeted retention strategies
• Improving service quality
• Building customer loyalty through personalized experiences

The project demonstrates that machine learning can effectively predict churn risk, allowing companies to take proactive measures in customer retention.